EP3879482A1 - Système et procédés de paiement de soins de santé basé sur la réussite - Google Patents

Système et procédés de paiement de soins de santé basé sur la réussite Download PDF

Info

Publication number
EP3879482A1
EP3879482A1 EP20020114.3A EP20020114A EP3879482A1 EP 3879482 A1 EP3879482 A1 EP 3879482A1 EP 20020114 A EP20020114 A EP 20020114A EP 3879482 A1 EP3879482 A1 EP 3879482A1
Authority
EP
European Patent Office
Prior art keywords
data
contract
outliers
health care
automated
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
EP20020114.3A
Other languages
German (de)
English (en)
Inventor
Leon Rebolledo
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Lyfegen Healthtech Ag
Original Assignee
Lyfegen Healthtech Ag
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Lyfegen Healthtech Ag filed Critical Lyfegen Healthtech Ag
Priority to EP20020114.3A priority Critical patent/EP3879482A1/fr
Publication of EP3879482A1 publication Critical patent/EP3879482A1/fr
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/22—Social work or social welfare, e.g. community support activities or counselling services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/10—Office automation; Time management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q2220/00—Business processing using cryptography
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L9/00—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols
    • H04L9/50—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols using hash chains, e.g. blockchains or hash trees

Definitions

  • the present invention relates in particular to the use of block chain technology in the healthcare environment.
  • EP 3 327 727 A2 a data processing system and method implementing an analytics platform and network information systems are known.
  • a healthcare resource plan is optimized based on a given time period, the risk level, the cost, the detection of patients at risk and so forth.
  • the clinical informatics platform is suggested to combine analytics systems for example disease specific analytic tools, predictive models and modules. It is suggested to extract the data, normalize the data, provide a robust clinical and networking analytics and deliver powerful and timely insights. It is stated that healthcare providers may be enabled to compare, analyze and identify best practices.
  • the data ingestion techniques may be applied to a heterogeneous system.
  • the system suggested may comprise tools for analytic model building, where inputs to the model in terms of rules, attributes, characteristics, criteria or the like may comprise diagnosis codes, test results, outpatient prescriptions and so forth.
  • a healthcare transactional validation file block chain is known. It is mentioned that enforcing privacy standards compliance, interoperability, data format conversion, ensuring proper treatment applied to the patient and maintaining a continuity of treatment records for individuals are issues in healthcare systems that manage large volumes of electronic medical records. It is suggested that healthcare transactions associated with the stakeholder are compiled into the chain of healthcare transaction blocks and that when the transaction is conducted, the corresponding healthcare parameters such as inputs and outputs, clinical evidence, outcomes etc. are sent to validation devices that establish the validity of the transaction and generate a new block in the block chain.
  • validation devices receive healthcare tokens that represent healthcare actions taken with respect to the stakeholder including for example test results for patient and diagnosis; the validation device shall also obtain an identifier of the respective historical block chain representing a chronicle of health care activities for the patient, doctor, insurance company etc.
  • Example healthcare tokens could include test results, a genetic sequence, a diagnosis, a prognosis, a patient identifier, a caregiver identifier and so forth.
  • the various service, systems, databases or interfaces exchange data are using standardized protocols or algorithms such as HTTP, HTTPS, AES, public-private key exchanges, web service APIs, known financial transaction protocols and that exchanges are conducted over for example the Internet, WAN, VPN or other type of packet switched network.
  • the cited document suggests that a common nomenclature ensures that transactions are processed in a uniform, repeatable and verifiable manner.
  • the document inter alia suggests that the peer obtains healthcare tokens and attempts to generate a validity block incorporating a validity token that could include simple information such as "agree/ disagree" or more complex information such as recommendations or alternatives diagnosis.
  • each block in the chain can be reviewed to analyze the path taken by a patient and that this would be of high value with respect to patients that have been prescribed similar, if not the same, drugs. Also, it is stated that healthcare transactions after the prescription is fulfilled can be analyzed to establish correlations of healthcare outcomes across patient populations.
  • the prior art also suggests to supply behavior analytics information as evidenced by historical healthcare block chain information to the various marketplaces such as Amazon, Google etc.
  • the marketplaces would then leverage such personal or demographic information to enable service providers to deliver highly targeted healthcare content to consumers, subject to privacy constrains.
  • medical data custodians are suggested to generate healthcare campaigns that are triggered based on features of the historical healthcare block chain information. It is suggested that access to the block chain can be offered to entities such as insurance companies, hospitals etc.
  • the insurance company could then apply actuarial algorithms to build risk tables without necessarily requiring patient identification of private information by analyzing sanitized data.
  • a system for authenticating records having an upstream network system with an upstream application in operative communication with the downstream application wherein the upstream application comprises a private block chain comprising at least one data record and wherein the private block chain has a plurality of nodes comprising at least one miner code that can validate certain data within the private block chain and authenticate the data record of private block chain such that the downstream application can detect and decrypt the authenticated data record.
  • the cognitive applications are suggested to be implemented so as to include inter alia healthcare services and financial services. It is also suggested to enable the management of a variety of data and metadata associated with various cognitive insight projects and to associate the data and metadata with billing information. It is noted that the use of real-time streams such as social media streams or device data streams is also suggested, for example medical devices tracking the patient's vital signs. The document suggests that patients could be informed if claims have been payed and an explanation of benefits be provided. It is suggested that payments could be initiated by smart contracts and that information provided to the patient and the provider may be structured to conform to confidential requirements such as the Health Insurance Portability and Accountability Act of 1996.
  • a private universal knowledge repository can be different from a public block chain knowledge repository and that knowledge elements stored in the private universal knowledge repository may be anonymized prior to being provided for inclusion in the public block chain knowledge repository.
  • an unsupervised learning machine may be implemented and that an unsupervised learning machine learning algorithm is not given a set of training examples, but attempts to summarize and explain key features of data processes, for example unlabeled data associated with the block chain.
  • a block chain associated cognitive insight may be provided to a healthcare insurance claims processor, recommending that an insurance claim associated with a particular claim be paid to a healthcare provider and that a process may be initiated informing a patient that the claimant has been paid and an explanation of benefit is provided.
  • US 2018/0203744 8 A1 relates to a data ingestion and analytics platform and notes that in generating insights from big data analytics, where a high variation in the incoming data rates may occur, the granularity of process data may vary, and high-availability and robustness are needed.
  • the data processing is suggested to comprise at least one processing layer split into plural lanes with an initial data ingestion layer operated for ingestion of data and routing said data to one of said plural lanes.
  • a personal care system employing block chain functionality is known. It is stated that it would be highly beneficial if multiple aspects of personal care could be recorded in an orderly manner and personalized care directed to the specific needs of an individual user could be provided.
  • a personal care device configured to maintain an interactive diary for the user is suggested and hardware personal care modules are suggested for assessing user healthcare needs and for providing healthcare information and user specific recommendations to the user.
  • a portable device is suggested to be taken with the user while traveling long-distance and connecting with the base device such that the user may record desired information and provide the information to the base location at a later time.
  • Examples given include the suggestion to assess the internal ear, determine a degree of hearing loss, assess buccal mucosa, tongue, mouth floor, teeth and gums, analyzing convex nail curvature, that is the angle between the nail and the nail bed, assess a swelling around the breast and armpit, ACNE, sleep/wake cycle and so forth. It is suggested that "health points" can be earned by users via a reward system for healthy behavior and that a user may redeem themselves for medical examination or medical related equipment or services. To this end, a block chain functionality is suggested. Furthermore, it is suggested that treatments can be subject to a contract where the actual payment will then depend on the outcome of the treatment relative to a result previously agreed upon.
  • US 2019/0027237 A1 relates to a block chain network for secure exchange of healthcare information.
  • the document assumes that electronic medical record systems if proprietary may impose heavy paperwork on physicians and staff, resulting in reduced patient quality of care while at the same time reducing opportunities for improving medical diagnosis and treatments by performing anonymized aggregate data analysis using machine learning or other techniques.
  • a user such as a physician may be authorized to access specific health records or be denied access.
  • each record may introduce unintended errors or the risk of fraud or record certification and that accordingly; electronic medical record systems must meet the needs of both healthcare providers and patients.
  • the document notes that block chain technology that can support any type of distributed application with centralized authority may be undesirable and that by using a distributed ledger of transactions, trust and integrity can be established without using third-party intermediaries.
  • the document thus suggests a server configured to receive a transaction from a client device via a publicly accessible network, wherein the server authenticates an identity of the client device prior to forwarding the transaction; and a private blockchain network comprising: a data aggregator node configured to receive the transaction forwarded from the server and distribute the transaction to at least a portion of a plurality of processing nodes, wherein each of the plurality of processing nodes executes a respective blockchain client configured to maintain a blockchain having contracts; and a forwarder node configured to, in response to detecting an event indicating that the transaction was successfully processed by a respective contract, transact protected health information in a data store separate from the blockchain.
  • WO2018/037148A1 a method and apparatus for block chain verification of a healthcare prescription is known. It is stated that patients visiting another country may have problems because a prescription valid in a first country may not be valid in a second country and because patients would not know which pharmacy could be trusted to give valid medicines. Furthermore, paper-based prescriptions are stated to be subject to forgery and that misuse may arise, such as fake doctors issuing prescriptions or real doctors prescribing too much or a single person getting too many prescriptions from multiple doctors. The document suggests using block chain technology to allow use of prescriptions across country boundaries, to save in costs required to build and maintain prescription databases and to minimize abuse possibilities that could cause death or support crimes.
  • a certified doctor creates a prescription and issues a transaction for the transfer of ownership of the prescription to a person, publishes the transaction in a block chain if this is allowed by the certificate of the doctor.
  • the transaction is to be "signed" by a government entity that approves the prescription and publishes this in the block chain.
  • the prescription can be transferred to a certified pharmacist by the patient, this transfer can be published and the patient can be provided with the medicine.
  • WO 2018/039312 A1 a block chain-based mechanism for secure health information resources exchange is known.
  • entities that store healthcare records containing protected health information must implement proper mechanisms for compliance to legal regulations and for managing them. It is also stated that different electronic health record systems are not interoperable with each other. It is suggested that a user such as a patient may be able, via an application executing on a client system, to set conditional permissions for the respective health information resource. These conditional permissions are suggested to be used to generate a permission that may be sent to a distributed ledger or healthcare block chain system to invoke an executable smart contract within a healthcare block chain.
  • neural networks such as convolutional neural networks or graph neural networks can be trained.
  • the two parties in a given country could base their agreement on data obtained in another country even though the population in the other country might have different lifestyle and eating habits, be subjected to more or less UV radiation, higher or lower ambient temperatures and so forth, all influencing the outcome of a medical treatment.
  • the lack of blood pressure reduction could be caused either by an error made when measuring the blood pressure or because in the peer group of the respective patient, the successful treatment simply is not related to reduction in blood pressure at all, for example due to a genetic predisposition locally predominant or due to a general physical constitution of the patient.
  • Machine learning systems usually are trained using the training data set in order to establish a machine model. Following the training, data can be input into the machine model in order to obtain the response of the machine to the input.
  • WO 2017/117 150 A1 Improving a machine learning system for creating and utilizing an assessment metric based on outcomes is suggested in WO 2017/117 150 A1 . It is stated that in healthcare, quality is currently measured using process metrics and customer satisfaction. The three most common outcome quality measures utilized are stated to be length of stay, readmission rates and mortality rates; the document observes that medical and pharmaceutical costs are only a portion of the true cost to manage a patient's condition and that a need exists to reduce the overall costs. It is suggested that the machine learning system obtains data from databases of providers, employees, clinical evaluators and employers. The clinical evaluator shall visit an employee initially to perform a physical and periodically thereafter, providing continuing input to a machine learning system, which allows the system to learn so that it can adjust quality factors based on changing conditions.
  • the system shall create a quality score and the scoring system shall be constantly adjusted and updated with both computer-generated data and human input from clinical evaluators.
  • episodic patients are analyzed based on procedures tied to a specific condition and duration and it is suggested to analyze claims where a rule was not satisfied in context with other claims to determine if such claims contributed to the overall cost of the episode. While this may help to improve the system, it will be understood that the learning system of the prior art is bothersome and expensive in view of the human evaluation necessary.
  • outliers and exceptions may constitute such new or unknown data.
  • outliers can come from many sources and hide in many dimensions.
  • the most common cause of outliers in a data set has been stated to be data entry errors (human errors), measurement errors (instrument errors), experimental errors (data extraction or experiment planning/executing errors), intentional errors (where dummy outliers are fed into a system to test detection methods), data processing errors (e.g. by incorrect data manipulation or unintended data set mutations), sampling errors (e.g. by extracting or mixing data from wrong or various sources), or the outliers are "natural".
  • Those "natural" outliers not due to intentional tests or errors are called novelties. Detecting novelties has already been the subject of considerable effort in data processing; in particular as detecting outliers can constitute a high computational burden.
  • novelties can be detected by a distance-based approach, where the point is considered a novelty if the distance to K-NN neighbors exceeds a predefined threshold; by an identity based approach where the data set is sparse, by clustering based, statistical approaches such as parametric approaches extreme values, hidden Markov models, hypothesis testing or nonparametric approaches.
  • a distance-based approach where the point is considered a novelty if the distance to K-NN neighbors exceeds a predefined threshold
  • identity based approach where the data set is sparse
  • clustering based statistical approaches such as parametric approaches extreme values, hidden Markov models, hypothesis testing or nonparametric approaches.
  • a clustering approach is discussed aiming to partition data into a number of clusters where to every data point, the degree of membership to each of the clusters can be assigned; novelty shall be detectable by comparing the degree of membership against thresholds to determine whether or not the data belongs to a cluster and novelty shall be given when a sample belongs to none of the available classes.
  • outliers are defined as "a case that does not follow the same model as the rest of data", including not only erroneous data but also "surprising veridical data".
  • the author suggests to derive from a database a decision tree which initially can be complex but fits a database well. Then, to make the complex decision tree simpler, an error associated with every node thereof is compared to the error obtained if the node would be pruned into a leaf. If the estimation error becomes smaller, the note is pruned and thus complexities reduced.
  • an automated data processing method comprising the computerized steps of training a machine learning model based on a first training set of data, evaluating new data with the trained model and identifying outliers, adding selected outliers to a training set and training the machine learning model with a second training set comprising the selected outliers.
  • the identification of outliers can be done in a fully automatic manner as computerized method even though in certain instances, parameters might be set by a human operator, for example defining an expected noise behavior of central signals such as the glucose level signals, blood pressure sensor signals and so forth.
  • outliers are not used blindly for training simply because they are outliers. Rather, only an outlier that has certain additional properties other than simply being an outlier shall be added to the training set for the reiterated training.
  • Such a selection could be fully automatic, for example by selecting a fixed fraction of all outliers that have been found to be not caused by errors, by selecting a group of outliers, where one or a plurality of parameters has but little variation; the selection could be semiautomatic for example in that suggestions are presented to human operator who has to confirm the selection; or the selection could be done fully manually.
  • the decision whether or not the selection is made automatically, semiautomatic, or manual can also be made automatically or by a human operator.
  • the selection of outliers will be done by a computerized step; also, in a preferred embodiment, the selection of outliers will be made in view of specific properties the selected outliers have.
  • outliers are not discarded or disregarded, but that rather the information related to the outsider is used as efficiently as possible to improve the initial model.
  • a very fast adaption of the machine learning model response to input data showing a slightly but noticeable different pattern compiled to the data pattern found in an initial training data set in lack of better data can be achieved. This helps to launch a machine-learning model significantly faster than before. Also, there is less need to "purify" data for a training set. It will be understood that this helps to reduce the cost or effort to launch a machine-learning model.
  • the first training set of data is obtained from a first source of data
  • the new data evaluated for identifying outliers comes from a second source of data
  • the second training set comprises both data from the first set and the selected outliers
  • the automated data processing method comprising the computerized steps of training a machine learning model based on a first training set of data obtained from a first source of data, evaluating data from a second source of data with the trained model and identifying outliers, adding selected outliers to a second training set and training the machine learning model with a second training set comprising the selected outliers.
  • the source of data to be evaluated is different from the source of data used for training. This is important because frequently, not only the sources of data are different, but also the pattern of the data coming from different sources will be different.
  • the present invention is mainly directed to medical healthcare systems, in particular data processing systems used in medical healthcare for determining success based payment, the selection of data points identified to relate to outliers and the addition of selected data points to the data training set in the machine learning model is helpful in a large number of machine learning model applications. It can be anticipated that a large number of cases exists where a training data set is not easily available prior to launching the system, for example because no real data had been available for training; i.e.
  • a manufacturer of cars may launch a car model where parameters such as fuel injection may be determined in response to a large number of sensor signals-however, the best response can only be determined in view of sufficiently large data sets which in turn are not available during the initial tests. Therefore, it might be possible to use as a first source data obtained with another car model or synthesized data and to then adapt the machine-learning model in view of real data. Also, the case may occur where tagging of data is impossible, for example in social media networks where new hashtags emerge, or where tagging is too expensive.
  • outliers are identified by at least one of a multivariate analysis, Chauvenet's criterion, a method according to ASTM E178 Standard Practice for Dealing with outlying Observations, Mahalanobis distance and leverage, Peirce's criterion, Tukey's fences, a distance based anomaly detection, a density based anomaly detection and a local outlier Factor. Explicit mention is also made of a Z score or Extreme Value Analysis.
  • outlier tests such as the Mahalanobis distance may be of particular use given that they can be used to identify unusual combinations of 2 or more variables. However, the average skilled person will also understand that while some of the outlier tests listed may be preferred, the list above is by no means exclusive.
  • data points may be outliers for a variety of reasons. It should be understood that those data points that are found to be outliers, that is that are identified as outliers, could thus be classified such that the classification either relates to at least one of a data entry error, a measurement instrument error, an intentional error, a data processing error, a sampling error or, alternatively relates to novelties.
  • the data points may constitute outliers because an error has occurred in measuring a blood pressure, in measuring a heart rate or in measuring the number of steps the patient has taken every day. If such errors are distinguished, it is in certain instances easy to see e.g. that all patients for whom an error has occurred in measuring the blood pressure have been treated in the same hospital and that correspondingly on the one hand, the hospital should be informed that the blood pressure data is incorrect and on the other hand, the correction of such error relating to a given patient group becomes feasible.
  • the classification of errors might improve both the acquisition of data and reevaluation of data acquired.
  • the machine-learning model it would be sufficient to just distinguish whether or not an outlier is due to an error at all.
  • the classification may be a simple flagging of outliers.
  • a machine-learning model is trained based on the collected outliers and based on those collected outliers that are best described by the model.
  • a sufficiently large number of outliers has been collected that do not relate to simple measurement errors or the like, but are due to the medical conditions observed in a sufficiently large group of patients, it is helpful to try to determine relevant pattern found in this group; then, if a pattern is found in the group of collected outliers, those outliers that correspond best to the pattern or to one of a plurality of patterns can be selected for addition to the training set of the machine learning model to be reiterated.
  • a different outcome in a different country may have a variety of reasons, for example, because in one country, patients are treated despite also suffering from another disease while in the other country, patients are denied treatment if they suffer from another disease; also, in different countries a significant portion of patients may have a significantly different diet, leading to a different response to treatment and so forth.
  • several groups of outliers should be identifiable provided the selection of samples is large enough. These groups could be identified using a machine-learning model on the collection of outliers. In this way, data points could be identified that clearly belong to each group. The data points found to belong to a group of outliers could then be selected for addition to the machine-learning model to be reiterated.
  • the number of outliers collected shall be considered large if the number is at least 1%, 2% 5%, 10%, preferably 15%, in particular preferred 20%, 25%, 30%, 33%, 50% or 75% of the number of entries in the training data set to which the outliers are to be added.
  • the number preferably is larger than 5%,10%, preferably 15%, in particular preferred 20%, 25%, 30%, 33%, 50% or 75% of the number of entries in the training data set.
  • the number is at least 15%, preferably 20%, 25%, 30%, 33%, 50% or 75% of the number of entries in the training data set. It would be understood that in a case where particularly strong deviations between the ensemble of data to be processed and the data ensemble constitute hearing the data training sets exist, the data training sets will not be a particularly good representation of the actual data to be processed; therefore, in case of particularly strong deviations, the impact and influence of the initial data training set only machine learning model should be reduced as soon as possible.
  • the plurality of iterations is done repeatedly, it is possible to reduce the number of outliers selected for addition to the training data set from iteration to iteration in the expectation that the training data set becomes better and better with each iteration. It is possible to simply add every outlier selected to an existing data training set or to exchange some of the data within the data training set against outliers selected. Exchanging data in the data training set against outliers allows keeping the computational expense of machine-learning constant for all iterations; adding data points to a given data set and thus increasing the size of the data training sets will increase the computational expense that may lead to better machine-learning models.
  • the outliers selected can either be added to the data set that is then split into two parts using standard procedures such as random selection within the data set, or, alternatively, some of the outliers could be added to the first part while other outliers selected are added to the control group.
  • the initial data set could rely on a random selection of patients in the first country A and the machine-learning model could be trained using this initial data set; then, to improve the evaluation of the success of the treatment in new country B, outliers observed in country B could be selected, but instead of adding them to a random selection of patients from just the first country A, data of random patients from the now larger number of countries A-E is selected for a new data training set and the outliers found in country B and selected for addition are thus added to this new random data set.
  • improvement of the machine-learning model is expected to be faster and the number of iterations necessary to reduce the number of outliers observed due to reasons other than errors or intentional erroneous test data is expected to be lower.
  • particularly strong deviations can be judged in some embodiments to be present if the overall percentage of patients for which a specific treatment works deviates by more than 5%, in particular more than 10%, in particular 15%, in particular 20%, 25%, 30%, 33%, 50% or 75% from the percentage of those patients that were considered in the data set for the initial or previous training of the machine learning models.
  • a particularly strong deviation can be judged to exist if the deviation has at least a 3-sigma significance, a 4-sigma significance or a 5-sigma significance.
  • a particularly strong deviation might be considered to exist if the absolute percentage of successful treatments differs by a large amount from the percentage in the training data set, for example by at least 5%, 10%, 15%, 20%, 25%, 30%, 33%, 50% or 75% percent, and, at the same time, the data training set is large enough so that the difference in the percentage is statistically highly relevant, in particular, having at least a 3-sigma significance, a 4-sigma significance or a 5-sigma significance.
  • an automated data processing method wherein data relating to a contract are processed and the contract has at least one condition agreed upon that can be fully or partially fulfilled, and the degree of contract fulfilment is automatically determined in view of the data processed relating to the degree of contract fulfilment.
  • the improvement in machine learning is not necessarily restricted to contract data and their evaluation.
  • the feedback provided by the data processing method of the present invention is applicable to a large number of applications.
  • any application that employs machine-learning techniques such as convolutional neural networks, graph neural networks and so forth that allow determining whether or not a given data point can be considered an outlier based on predefined statistical methods may profit from the invention and the feedback suggested herein.
  • the present invention achieves a faster improvement of the system response in all technical areas with machine learning and can be used in a variety of areas such as control of automobiles, laundry machines, production machines, dishwashers and so forth.
  • the invention also is of particular advantage for contracts in general, and smart contracts in particular, as the selection of outliers identified and the addition thereof to a training data set will help to improve the response of the machine learning model particularly quickly, so that where conflicting interests of different parties need to be taken into account, any initial imprecise response will not persist for a very long time and will thus be more easily acceptable to all parties involved, which in turn allows to meet an agreement faster.
  • an automated data processing method wherein data related to a contract are processed and the contract has at least one condition agreed upon that can be fully or partially fulfilled, and the degree of contract fulfilment can be automatically determined in view of the data processed relating to the degree of contract fulfilment
  • the method comprises the computerized steps of ingesting data from a plurality of sources, automatically determining an extent to which the contract has been fulfilled in view of the ingested data and in view of the at least one condition agreed upon in the contract, wherein an extent, to which the contract has been fulfilled, is automatically determined by computerized steps in response to a machine-learning model trained based on a first training set of data, and outliers are identified by a statistical analysis, and are evaluated using computerized steps, the automated method further comprising execution of at least one of the steps of amending the machine learning model by training the model with a second data set comprising the outliers, amending the extent of contract fulfillment determined, suggesting an amendment to the contract, automatically amending the contract in view of outliers selected according
  • the automated data processing method of the embodiment may or may not have all features described to be present in preferred embodiments.
  • an automated contract data processing method wherein the contract is a smart contract, the smart contract has at least one condition agreed upon that can be fully or partially fulfilled, data is available relating to the degree of contract fulfilment, and the degree of contract fulfilment can be automatically processed, the method comprising the computerized steps of ingesting data automatically obtained from a plurality of sources in accordance with the smart contract, automatically determining an extent to which the contract has been fulfilled in view of the ingested data and in view of the at least one condition agreed upon in the contract, wherein an extent, to which the contract has been fulfilled is automatically determined by computerized steps in response to a machine learning model trained based on a first training set of data, and outliers are identified by a statistical analysis, and are evaluated using computerized steps, the automated method further comprising execution of at least one of the steps of amending the machine-learning model by training the model with a second data set comprising the outliers, amending the extent of contract fulfillment determined, suggesting an amendment to the contract and/or automatically amending the contract in view
  • an initial agreement in a contract contains provisions relating to conditions under which the contract can be amended and/or the manner in which the contract is determining results such as payments to be made
  • meeting amended conditions and changing the manner results are determined can be thought of as "amending the contract" even though the actual contract has not been amended, but only the conditions under which for example payments are to be effected have been adapted to an improved knowledge or database as agreed upfront.
  • smart contracts are particularly helpful where data need to be provided to other parties as the smart contracts may be implemented such that the communication of the data relevant to the contract can be effected automatically. It will be understood that this is particularly helpful where data needs to be obtained from a large number of parties, for example a large number of hospitals in any one country.
  • the contract is a smart contract relating to treatments by health care providers and health care payers
  • the smart contract has at least one condition agreed upon between a health care provider and a health care payer that can be fully or partially fulfilled
  • data are available relating to the degree of contract fulfilment
  • the degree of contract fulfilment can be automatically processed
  • the method comprising the computerized steps of ingesting data automatically obtained from a plurality of health care providers in accordance with the smart contract, automatically determining an extent to which the contract has been fulfilled in view of the ingested data and in view of the at least one condition agreed upon in the contract, wherein a success-based payment relating to the extent to which the contract has been fulfilled is automatically determined by computerized steps in response to a machine learning model trained based on a first training set of data obtained from a first health care provider or a first group of health care providers, and outliers are identified by a statistical analysis, and are evaluated using computerized steps, the outliers being obtained from one or more second health care providers different from the first health
  • the use of the invention in healthcare systems is of particular advantage. This is due to the fact that on the one hand, a situation is frequently given where seemingly small differences from patient group to patient group may have a significant impact on the outcome of the treatment. Thus, it frequently is necessary to not only retrain the machine-learning model in view of an enlarged data set available after some time of data ingestion, in particular after some time of data ingestion from a second, different source, but also to improve the machine learning model as soon as possible to best match any current data set and to thoroughly take into account new insights available.
  • the contract is a block-chain based smart contract relating to treatments of patients by health care providers and to success-based payments by health care payers.
  • the data that is automatically obtained from a plurality of health care providers in accordance with the smart contract may be subject to confidentiality regulations and ingesting data is done by an intermediate party different from both the health care providers and the payers.
  • the use of block chain methods is preferred because in this manner, even highly sensitive data can be transferred without the risk of being intercepted. Also, trust between parties involved such as a payer of healthcare measures and the provider of healthcare is automatically established when using block chain methods.
  • the communication of data obtained from the sources is preferably based on block chain technology
  • the data may be stored for analysis outside the block chain. It will be understood that this facilitates the retraining of models and that additional handling of the data is simplified, for example for synthesizing anonymized data from the entirety of data ingested.
  • the data based on the ingested data is thus stored by the intermediate party outside of the block chain.
  • the data ingested by the intermediate party is extracted, loaded and stored outside the block chain in an untransformed manner.
  • This data might then be used for purposes other than just for the fulfilment or evaluation of a smart contract.
  • synthetic data for research purposes. It will be understood by a skilled person that referring to storing data in an untransformed manner in a preferred embodiment means that what is stored is the raw data; this is preferred because error detection based on raw data is simpler. For example, where certain signals constituting part of the data are very small prior to a normalization step, it will be easily understood that such small signals will be prone to noise while any normalized data will not be related to the influence noise has on small signals. Thus, where a small signal is strongly influenced by noise, storing data in a normalized manner might prevent a user from understanding the impact of noise on given data points.
  • an automated health care contract data processing method as described above is suggested wherein healthcare contract data is processed, at least one health care provider collects data from individual patients or from a plurality of connected partners and the data ingested by the intermediate party is the data collected by the least one health care provider.
  • a hierarchy is provided to reduce the burden on an intermediary party running machine-learning models for contract evaluation or at least ingesting large amounts of data.
  • the data will be needed locally as well, for example, where a patient transmits data relating to his health condition (such as blood pressure data, heart rate data and insulin levels) to a hospital following a medical intervention such as surgery.
  • a medical doctor in the hospital might supervise the progress of the patient by reviewing the data or by receiving notifications relating to anomalies so that the treatment can be adapted in view of the recovery of the patient or of complications identified.
  • the data from a healthcare provider such as a large hospital can be transferred to the intermediate party in a block-wise manner such that information relating to a large number of patients is transmitted in only one file or in a small number of files.
  • a healthcare provider such as a large hospital
  • an automated health care contract data processing method is preferred wherein at least one health care provider collects data from individual patients or from a plurality of connected partners and the data ingested by the intermediate party is the data collected by the least one health care provider and the data obtained and is transferred in response to a number of requests smaller than the number of individuals the data relates to.
  • a system is shown that can be used in an automated data processing method comprising the computerized steps of training a machine learning model based on a first training set of data, evaluating new data with the trained model and identifying outliers, adding selected outliers to a training set and training the machine learning model with a second training set comprising the selected outliers.
  • figure 1 serves to show on its left side some of the hardware necessary for healthcare providers. On its right side, figure 1 symbolizes infrastructure and its use by a service payer such as a health insurance company. In between the healthcare provider side and the service payer side, a diagrammatic scheme for explaining the way the outcome of healthcare services provided by the health care provider is assessed such that a success-based payment to be made by the payer to the healthcare provider can be determined is shown.
  • an electronic health record EHR is maintained by the provider.
  • the electronic health records of patients will be managed and stored on a server operated by the healthcare service provider, for example a central hospital server.
  • a server operated by the healthcare service provider for example a central hospital server.
  • other possibilities exist instead for storing EHR data other than on a hospital server e.g. a "software as a service" (SaaS-) EHR provider or storing data in a data cloud. While the invention could also be used in such cases, it will be discussed hereinafter context with a hospital server.
  • SaaS- software as a service
  • Some of the data compiled into the electronic health records may be entered directly by a human, for example a nurse or a physician; this case is indicated in figure 1 . It will be understood that specific applications may be provided to help a physician, nurse or other involved person to enter data into the electronic health record; this holds in particular where data hitherto not of interest needs to be compiled into the electronic health record.
  • the corresponding app can be defined.
  • Data may also be entered automatically into the electronic health record, for example by electronically sending the results of laboratory examinations to the server after values such as hemoglobin blood levels or the like have been determined in a medical laboratory.
  • the server may also communicate with an app running on a smart phone of a patient to allow input of patient reported outcome measures and patient reported experience measures (conventionally termed PROM/PREM" response).
  • data compiled into the electronic health record may also comprise patient reported measures as well as data directly acquired using for example a patient's smart phone; as an example, the smart phone could be used to record the number of steps the patient walks every day.
  • patient reported measures are evaluated, it is taken into account that these measures typically are highly subjective. Nonetheless, entering the PROM/PREM response into the electronic health record kept by the healthcare service provider or correlating the subjective feeling of well-being as expressed by the responses to the current treatment may be important to a physician and should thus be made available to the physician.
  • the outcome of the success of the treatment can be assessed even though the PROM/PREM response of the patient is not entered into the electronic health record and that a possibility exists to directly transfer patient responses or other data acquired by a patient app on a patient's smart phone, for example data indicators of the distance the patient has walked at any given day, to an institution intermediate between the healthcare service provider and the healthcare service payer. Also, it should be noted that a possibility exists to directly transfer part of a patient response or other data acquired by a patient app on a patient's smart phone, for example data indicators of the distance the patient has walked at any given day, to only an institution intermediate between the healthcare service provider and the healthcare service payer. This is particularly helpful in cases where continuous or repeated treatment in a hospital is not necessary.
  • a patient response may be a binary response ("do you have migraine today Y/N") or a non-binary response ("how severe are you affected by migraine today on a scale of 1 to 10? "). In this manner, it becomes possible to for example only partially reimburse the cost for a drug if a patient has suffered more from migraine than promised by the manufacturer of the drug.
  • the electronic health record comprises information that a local or national body of physicians has agreed to be useful in describing the health status of patients suffering from a specific disease and in determining the way a patient needs to be treated.
  • a situation may occur where physicians in a given hospital and/or in a given country are not sufficiently aware that certain other parameters have been found important in other hospitals or countries, for example levels of specific trace elements found in the blood, parameters such as glucose levels that can point to metabolic dysfunctions such as diabetes and so forth.
  • the assessment of the outcome of a treatment has gained importance in the recent past and that accordingly, for a number of diseases, standardized expected outcomes to which parties involved can easily agree have been defined, compare for example the references found at the date of application at www.ichom.org.
  • rules can be set in an initialization phase of a smart contract so that the correct information is gathered; in particular, it is possible to set these rules such that data are entered into the electronic health record in a manner easily assessable.
  • a rule can define how often certain parameters need to be measured and in what way they are measured, for example defining that blood glucose levels must be determined after a period of at least 6 hours of fasting.
  • a treatment achieves an outcome agreed upon before, in the context of the present invention it will be considered successful or "a success". Accordingly, it is not necessary to completely restore health of a patient or to completely eradicate every symptom of a disease in order for a treatment to be successful. Also, situations may arise where the objectives of the treatment are not fully achieved; for example, where a patient suffers approximately 4 out of 7 days from migraine without drugs, a treatment might aim to keep the patient free of pain from migraine attacks for at least 25 days out of a 4-week-period. Where the patient is pain-free only 24 days out of the 4-week-period, the outcome is not fully met.
  • data can be transmitted from the medical app of a patient device to the server or other hardware keeping the electronic health record by either pushing or pulling.
  • a predefined transmission scheme is executed, in particular as agreed between partners to the smart contract.
  • pushing one or more update events of the electronic health record database and/or the patient app database may trigger transmission of data to the electronic health record. It should be noted that it is not necessary to push data into the electronic health record with every update event, in particular not in case that the update event does not justify urgent intervention by a physician.
  • a decision whether or not new data entries into the electronic health record should be pushed can be made simultaneously depending on a plurality of parameters such as the time since the last transmission, a judgment on the importance of new entries, the amount of data to be transmitted and the availability of suitable transmission channels such as Wifi, GSM, G4 or the like. Also, it should be noted that a large number of cases would exist where intervention by a hospital, a physician or the like will not be needed nor should specifically be prepared for, as is for example the case for pain treatment, treatment of chronic diseases or states such as obesity and the like.
  • an interface is provided defining the information in the electronic health record needed.
  • the system is applicable in particular where the treatment uses specific drugs in a treatment and a determination has to be made to what degree a pharmaceutical company producing the specific drugs is to be reimbursed.
  • the success based payment or part thereof may go to the pharmaceutical company rather than a hospital in which the drugs are administered.
  • a possibility would in particular exist to reimburse a hospital for the work done there, but to not reimburse the pharmaceutical company if a drug administered according to predefined schemes does not yield any success.
  • the pharmaceutical company could be seen as a "health care provider", generally, in the present description and claims, reference will typically be made to only a hospital, a physician or the like for simplicity. It will also be understood that under certain circumstances, payments can be made during the course of the treatment according to a smart contract, and, in case the treatment should turn out to be not successful, as can be determined that all or some of the payments need to be paid back. This will be considered a “reimbursement" in the context of the present invention as well.
  • an interface is defined according to which certain information in accordance with the smart contract can be requested from electronic health record.
  • FHIR API such as the HAPI FHIR API; these abbreviations stand for Healthy Americans Private Insurance Plans - Fast Healthcare Interoperability Resources- education programs interface.
  • a suitable interface is agreed upon according to the present invention.
  • UMLS Unified Medical Language System
  • a health service provider may request confirmation that a specific patient is eligible to participation in a certain treatment program, such as being eligible to receiving expensive cancer drugs or a new drug to reduce blood pressure. Such request may come up during the treatment of a patient already participating in other treatment programs; also, a case might arise where a patient is to be treated using drugs that have been not yet been marketed in the country of treatment before. Where eligibility is to be assessed, the program interface might automatically gather information whether (and if so, what) additional or alternative treatments are requested for a specific patient. It will be understood however that a general request can be sent to the server keeping the electronic health records of a plurality of patients to indicate whether agreements relating to new patients should be met.
  • the respective information can be retrieved from the respective data storages of for example the smart phone of the patient or the electronic health record in a clinic or the practice of a physician and a success-based payment can be determined. This can be done following the flow diagram shown in figure 2 .
  • the data communicated may also comprise other non-medical information, such as geographical information, general ambient temperatures during a period of treatment, or data that allows to judge whether or not the patient has adhered to conditions agreed upon in the contract such as walking a minimum number of steps per day or reducing the intake of calories.
  • connection to these devices in particular a direct connection, for example via API/Web services based on either FHIR or custom built or other similar in nature.
  • a more automated communication which is particularly useful for assessing the outcomes of a large number of treatments from a large number of hospitals in a plurality of countries. For example, the outcome of different treatments, for example at least 5, preferably at least 10 different treatments, for a number of patients exceeding 50 patients, preferably exceeding 100 patients each, the patients coming from at least 5, preferably at least 10 different hospitals, in particular in at least 3, preferably at least 5 different countries.
  • a fully automated system using a messaging system for transferring data is preferred.
  • the messaging could be based for example on XML, flat file or JSON format or others. Accordingly, once the automated process of transferring data has been established, there is no need to remind any person to retrieve the data nor is there any need to spend any working hours effecting transmission. It will be understood that where the automated transfer is established as is preferred, the data transmitted will follow a data retrieval protocol agreed upon in the smart contract that all data necessary to assess the outcome of a treatment as well as any additional contract and/or adherence conditions and/or, depending on requirements, certain other segments of the FHIR resources or others relevant for determining the payment.
  • the data itself is subjected to a determination whether or not the data is corrupted and/or whether other errors are present in the data. Errors could be corrupted due to a hardware failure and errors could occur for example due to mistakes by a physician, nurse or patient when entering the raw data. Also, a possibility might exist that data has been intentionally manipulated.
  • a check for consistency might rely on each single data entry alone, for example evaluating whether it is reasonable to assume that an allegedly very significant weight loss has occurred in a very short time; also, a consistency check could be carried out using a plurality of data; for example, it is unlikely that a patient states in a patient response to feel very good while at the same time high fever is recorded. Where data cannot be directly accepted, it could either be rejected where it is obvious that a data entry has been erroneous or has been corrupted, or it can be flagged and a request or message can be sent to the source of data such as the electronic health record, patient or the like to verify or correct the data.
  • the data will be assessed by first extracting the relevant data from the blockchain and the extracted data is collected in a repository termed "data lake” from which it can be easily retrieved without further decryption or the like. It will be understood that the data lake will be strongly protected against intrusion using firewalls and other techniques known per se.
  • the ingestion of data into the data lake will comprise steps such as normalization, for example using Celsius instead of Fahrenheit for all temperatures.
  • the outcome of a treatment is determined by feeding the data into a suitable artificial intelligence model.
  • a machine-learning model is used; the present invention also suggests identifying outliers and selecting certain outliers so that the model can be improved.
  • outliers are of particular interest where one or more parameters can be identified that have not been regarded to be important in an initial model but now a indicate a parameter pattern previously disregarded should be considered.
  • the parameters that are relevant for an initial model will only be known once the model has been trained. This implies that more data needs to be ingested into the data lake than later on necessary to assess the success of a treatment based on the initial model. Therefore, reference is made in figure 2 to "pruning". It will be understood that while the additional data significantly helps to select relevant outliers, the additional data need not be evaluated in the initial model.
  • cases patients might not have adhered to conditions agreed upon in the contract. This might be the case because of the patient has not bothered to stick to a medication regime or has not bothered to keep a strict diet; in such a case there could be provisions in the insurance contract of the patient allowing the insurance company to decline payments and therefore, no payments would have to be made to the hospital by the insurance company (although, obviously, payments would then have to be made by the patient in the case mentioned).
  • treatment of a patient will not be reimbursed unless there has been a recorded adverse event or complication that indicated that clinically a dose reduction or a change of frequency (interruption or discontinuation) has been prescribed by a physician such as an oncologist.
  • the data is then analyzed using the initial machine-learning model and payments to be effective according to a contract agreed upon between the healthcare providers and the healthcare payers are determined. This is done in a conventional manner. However, in doing so in a number of cases, no payment or only very little payment is to be effected for a variety of reasons for example because severe side effects have occurred in the range of patients or because the number of patients did not react to the treatment as predicted and expected, although some improvement was still observed.
  • the invention now suggests to analyze whether in the data there is any systematic pattern to be found that is not reflected in the machine-learning model; accordingly, in particular the data of those patients that have been treated without full success, in particular in a manner where no payment must be made according to the contract initially agreed upon, should be analyzed.
  • the respective data points can be treated as outliers.
  • outliers are not restricted to patients that have been treated without success. For example, where a drug is administered to patients to reduce virus load, for example the HIV-load found in AIDS patients, payments could be effected if the HIV load is below a certain non--zero threshold. Among those patients having a significantly reduced HIV load, there might be some patients where the load found is exactly zero (that is, below the detection limits).
  • a multivariate analysis can be executed, using in particular one of Chauvenet's criterion, a method according to ASTM E178 Standard Practice for Dealing with outlying Observations, Mahalanobis distance and leverage, Peirce's criterion, Tukey's fences, a distance based anomaly detection, a density based anomaly detection and a local outlier Factor. In this manner, those outliers that exhibit a common behavior can be identified. In view of this, and figure 2 , References had to multivariate data analytics, MVDA.
  • outliers that relate to treatments that have not been as successful as expected, in particular outliers that relate to treatments that have missed success only by a small margin, warrant further analysis to determine whether amendments or additions to the initial smart contract and the agreements made therein are justified. For example, a case might occur where treatments are evaluated in view of data gathered in other countries and several local factors contribute to the success or failure of a treatment such as average ambient temperatures, different general eating habits and the like.
  • different patterns will exist for the different countries and the present invention allows adjusting contracts to different patterns identified much faster than before. The actual manner how to deal with such patterns emerging from the analysis of outliers can depend on the exact agreement met.
  • a "new knowledge" - provision can be included in the initial contract pre-defining how parties to the contract shall deal with groups of outliers relating to a newly formed common reason for over-performing or slightly underperforming. If such a provision is included in the initial contract, no amendment to the contract itself is needed at all, although in most cases a notification should be sent to all partners involved. In other cases, where partners prefer to leave a contract as is once it has been agreed upon, a suggestion can be made, in particular be made automatically, to meet a child contract reflecting the slightly different response of patients of the outlier group to a treatment and accordingly defining success and/or conditions, in particular adherence conditions, in a manner different from the provisions in the original (parent) contract.
  • the previously used training set could for example be enlarged by adding the selected outliers, by replacing a fixed number of previously included set elements by selected outliers or by adding the outliers to a base set from which a training set is (randomly) chosen for training a new model.
  • either the initial assessment of treatment success is maintained and/or data are additionally assessed according to a child contract and/or, where both a common pattern has been found in a sufficiently large plurality of outliers such that a retraining of a machine learning model is justified and the initial agreement contains provisions to automatically amend the conditions for payment, a new assessment can be effected and corrected success -based payments can be calculated.
  • the calculated results need to be communicated to the parties involved.
  • the healthcare provider and the healthcare insurance company usually are informed; in selected cases, the patient could also be informed.
  • the data can be maintained as is or can be anonymized so that at a later stage, the information contained the overall set of data collected can be made available for research.
  • a first check is made for every patient whether or not the patient is part of the program that is whether the medical data of the patient need to be evaluated. It will be understood that a large number of patients in a hospital will not receive any relevant treatment and thus health records should be skipped. For example, the patient may suffer from an entirely different disease.
  • a check is made whether the patient is eligible for a success-based payment according to the smart contract agreed upon.
  • the patient must satisfy certain conditions, for example with respect to his age, health insurance, disease-for example where cancer is concerned the stage of the cancer, the histology, presence or absence of mental status in certain organs and the like. It can all be also be judged whether the patient suffers from the specific form of cancer for a very long time already or whether he has been newly diagnosed. Patients that have already suffered from a specific disease such as cancer for very long time may be significantly more difficult to treat; in particular, where a risk exists that metastasis occurs. Accordingly, the judgment is made to exclude patients that have suffered from the specific disease for too long.
  • the patient is eligible to the program since he has been diagnosed only recently. Accordingly, then, the patient may be treated in a manner reimbursing a pharmaceutical company for delivering a personalized drug only in case of success and a corresponding treatment plan defining exact delivery of pharmaceuticals both with respect to timing and dosage is established.
  • problems might have shown early on, so that an initial treatment plan has been abandoned completely and a new line of treatment has been started; accordingly, the initial treatment line would not be "open" any longer.
  • the patient will be rejected from the program and no further data concerning this patient will be acquired for the time being.
  • a physician to re-enter a patient into the program and that then, the next time the electronic health records of the hospital are accessed, a new evaluation could be detected. For example, obese patients could be considered to be not eligible, and a patient who initially during a first evaluation had the body mass index beyond the threshold accepted by the smart contract has lost significant weight ever since; in that case, the patient could be reintroduced to the contract and the data be read anew.
  • Eligibility relies on a number of tests and conditions. In order to avoid that once the patient is considered eligible, the treatment is only started over after a very prolonged time-resulting in a worsening of condition and the beginning of treatment, a check is then made that the assessment is not too old. Where the assessment for eligibility is too old without any planned treatment, the patient is rejected from the program (could be reentered once the tests have been updated).
  • the next check is made is whether or not the therapy that is to be paid in a success-based manner according to the contract is going on or not, that is whether or not the therapy line is active.
  • the therapy line is not active, obviously, the patient is not treated any longer; in such a case, there cannot be a success based payment as not all conditions agreed upon in the smart contract has been adhered to. Otherwise, that is when the therapy line is active, a check is made whether the full treatment has already been given or not.
  • data will be collected indicating whether all medicaments have been administered as planned.
  • data relating to complications that might have occurred are collected together with data indicating whether in view of complications a reduction of the doses of medicaments or complete discontinuation of treatment has been prescribed by a physician treating the patient.
  • a plurality of different drugs are administered and the dates of administration together with the dosages of each of the pharmaceuticals administered are recorded in the electronic health record. Then, further data is collected that allows the assessment of the treatment, for example relating to the well-being of the patient, the detection of further metastasis, a reduction of size of existing tumors and so forth.
  • the data collected in the electronic health record is stored in a server of the hospital where the patient is treated.
  • a program is periodically running to automatically generate blockchain blocks including the information gathered for those patients that at the time of generation of blocks are part of a success-based treatment plan.
  • the blockchain to which the blocks have been added is then distributed to all parties involved, in particular an intermediate assessing the success of treatments.
  • This intermediate party will then determine at some point whether all medicaments have been administered and treatment has been given as planned and, if this will not the case, not whether any complications have occurred that justify a deviation from the preagreed treatment plan. If so, it is checked whether a dose reduction or discontinuation of certain pharmaceuticals has been prescribed by the oncologist responsible for treating the patient.
  • the check is made whether all conditions agreed upon have been complied to. For example, a check can be made whether or not the drugs that have been given in the prescribed intervals, with the correct doses and so forth. It will be understood that in a case where a physician has decided to change the dose of one of the drugs the corresponding entry the electronic health record is found, adherence is given if the doses administered correspond to the changed doses. Thus, where a deviation from the treatment plan has been recorded but no complications have occurred, the patient is excluded from the assessment of his success and the parties are informed that payments must be determined differently.
  • a check is made whether or not sufficient time has passed since the last treatment.
  • the treatment will be considered successful only if no matter status have been observed a sufficiently long time after treatment has ended. Therefore, the check is made whether or not sufficient time has passed.
  • the time of 204 days is considered the minimum time since the end of treatment.
  • the intermediate party can determine in view of the data whether or not the state of the patient shows that the disease has progressed, for example if metastatic tumors have been observed, existing tumors have grown and so forth.
  • the patient obviously does not respond in the expected manner to the treatment and will be excluded from the program in the future.
  • the payment for treatments effected thus far needs to be determined; such payment could for example reimburse the hospital for treating the patient, examining samples in laboratories etc.. However, payments to the pharmaceutical company producing the drugs could be denied.
  • the decision is made what other liabilities exist and any reimbursements necessary that are not success-based are calculated. For example, a payment is made to the hospital but not to the manufacturer of drugs.
  • the patients are kept in the program so that additional data can be acquired during subsequent data accessing cycles (as agreed upon in the contract) and the decision later on is made whether or not these patients are to be treated as relevant outliers.
  • outliers may be selected from data where complications have occurred and dosage had to be adapted and/or in cases for which during the calculations it is determined that the success of treatment does not justify full payments or does not even justify partial payments.
  • outliers may be selected from data where complications have occurred and dosage had to be adapted and/or in cases for which during the calculations it is determined that the success of treatment does not justify full payments or does not even justify partial payments.
  • success-based payments can be reliably determined in a secure manner. Furthermore, the determination of what a fair success based payment not only is simple; also, it will be understood from the above disclosure, that the process of such determination is transparent and reliable, so that parties to the contract can be expected to agree on the conditions of a contract faster. This holds in particular as the disclosed identification and use of outliers helps to improve the conditions of a contract in view of knowledge gained since the instantiation of a contract.
  • the present invention not only allows improvements with respect to success-based smart contracts relating to a large groups of patients; rather, it is also possible to use the technique of the present invention and in particular the fact that knowledge about success to be expected can be used very fast after acquisition.
  • This allows to not only determine conditions of a success based smart contract for large groups of patients, but will also help in a flexible dynamic pricing for specific, small groups such as a rather small number of patients having a number of well-defined comorbidities.
  • the way outliers are handled according to the present invention allows to very easily to determine fair conditions for success based payments even though in a more conventional set up, such patients with comorbidities would most likely to present a high-risk of treatment failure and/or be considered to be hardly treatable successful.
  • the invention also allows to factor in a larger number of parameters, resulting for example in a flexible or dynamic pricing using for example an algorithm embedded in a "smart" contract initially agreed upon depending on parameters such as the gender of the patient, the ethnicity of a patient, the age of the patient, measures of physical activities of a patient, comorbidities of the patient, treatment methods of comorbidities.
  • outliers are used according to the invention in a specific manner, namely to improve the machine learning models in a specific manner. While this has been explained with respect to processing smart contract data, improving the machine learning models based on outliers identified vis-à-vis an initial model is helpful in other fields as well.
  • an initial model could be based on the average behavior of a given car type having aggregates controlled in view of a plurality of parameters according to a machine learning model.
  • the sensor response and/or the aggregate response in a specific automobile of the car type might deviate from the average behavior, leading to clusters of relevant outliers.
  • the outlier treatment of the present invention might lead to improved behavior of the car type. It will be understood that similar advantages might be obtained in another technical fields.

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Tourism & Hospitality (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Marketing (AREA)
  • Economics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Child & Adolescent Psychology (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
EP20020114.3A 2020-03-09 2020-03-09 Système et procédés de paiement de soins de santé basé sur la réussite Pending EP3879482A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP20020114.3A EP3879482A1 (fr) 2020-03-09 2020-03-09 Système et procédés de paiement de soins de santé basé sur la réussite

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
EP20020114.3A EP3879482A1 (fr) 2020-03-09 2020-03-09 Système et procédés de paiement de soins de santé basé sur la réussite

Publications (1)

Publication Number Publication Date
EP3879482A1 true EP3879482A1 (fr) 2021-09-15

Family

ID=69804369

Family Applications (1)

Application Number Title Priority Date Filing Date
EP20020114.3A Pending EP3879482A1 (fr) 2020-03-09 2020-03-09 Système et procédés de paiement de soins de santé basé sur la réussite

Country Status (1)

Country Link
EP (1) EP3879482A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230130086A1 (en) * 2021-10-27 2023-04-27 Arm Limited System, devices and/or processes for modifying electronic contracts

Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2002059746A1 (fr) 2001-01-26 2002-08-01 Whizbang Labs Procede d'apprentissage et de combinaison de regularites locales et globales pour extraction et classification d'informations
WO2017117150A1 (fr) 2015-12-28 2017-07-06 Integer Health Technologies, LLC Système d'apprentissage machine pour créer et utiliser des mesures d'évaluation sur la base de résultats
US20180048461A1 (en) 2016-08-10 2018-02-15 Peer Ledger Inc. Apparatus, system, and methods for a blockchain identity translator
WO2018039312A1 (fr) 2016-08-23 2018-03-01 BBM Health LLC Mécanismes à base de chaînes de blocs pour l'échange sécurisé de ressources d'informations de santé
WO2018037148A1 (fr) 2016-08-22 2018-03-01 Nokia Technologies Oy Procédé et appareil permettant une vérification de chaîne de blocs de prescriptions médicales
US20180103042A1 (en) 2016-10-12 2018-04-12 Bank Of America Corporation Automated data authentication and service authorization via cryptographic keys in a private blockchain
EP3327727A2 (fr) 2016-11-23 2018-05-30 Optum, Inc. Systèmes et procédés de traitement de données mettant en uvre une plate-forme analytique améliorée et des systèmes d'informations en réseau
WO2018098077A1 (fr) 2016-11-23 2018-05-31 General Electric Company Systèmes médicaux à apprentissage profond et procédés associés à des procédures médicales
US20180165416A1 (en) 2016-12-09 2018-06-14 Cognitive Scale, Inc. Method for Providing Healthcare-Related, Blockchain-Associated Cognitive Insights Using Blockchains
US20180203744A1 (en) 2017-01-09 2018-07-19 Alexander WIESMAIER Data ingestion and analytics platform with systems, methods and computer program products useful in conjunction therewith
US20180248981A1 (en) 2013-11-14 2018-08-30 Mores, Inc. Enhanced personal care system employing blockchain functionality
US20190027237A1 (en) 2017-07-21 2019-01-24 Patientory, Inc. Blockchain network for secure exchange of healthcare information
US10340038B2 (en) 2014-05-13 2019-07-02 Nant Holdings Ip, Llc Healthcare transaction validation via blockchain, systems and methods
US20190379642A1 (en) * 2018-06-08 2019-12-12 Gcp Ip Holdings I, Llc Blockchain Overwatch

Patent Citations (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2002059746A1 (fr) 2001-01-26 2002-08-01 Whizbang Labs Procede d'apprentissage et de combinaison de regularites locales et globales pour extraction et classification d'informations
US20180248981A1 (en) 2013-11-14 2018-08-30 Mores, Inc. Enhanced personal care system employing blockchain functionality
US10340038B2 (en) 2014-05-13 2019-07-02 Nant Holdings Ip, Llc Healthcare transaction validation via blockchain, systems and methods
WO2017117150A1 (fr) 2015-12-28 2017-07-06 Integer Health Technologies, LLC Système d'apprentissage machine pour créer et utiliser des mesures d'évaluation sur la base de résultats
US20180048461A1 (en) 2016-08-10 2018-02-15 Peer Ledger Inc. Apparatus, system, and methods for a blockchain identity translator
WO2018037148A1 (fr) 2016-08-22 2018-03-01 Nokia Technologies Oy Procédé et appareil permettant une vérification de chaîne de blocs de prescriptions médicales
WO2018039312A1 (fr) 2016-08-23 2018-03-01 BBM Health LLC Mécanismes à base de chaînes de blocs pour l'échange sécurisé de ressources d'informations de santé
US20180103042A1 (en) 2016-10-12 2018-04-12 Bank Of America Corporation Automated data authentication and service authorization via cryptographic keys in a private blockchain
EP3327727A2 (fr) 2016-11-23 2018-05-30 Optum, Inc. Systèmes et procédés de traitement de données mettant en uvre une plate-forme analytique améliorée et des systèmes d'informations en réseau
WO2018098077A1 (fr) 2016-11-23 2018-05-31 General Electric Company Systèmes médicaux à apprentissage profond et procédés associés à des procédures médicales
US20180165416A1 (en) 2016-12-09 2018-06-14 Cognitive Scale, Inc. Method for Providing Healthcare-Related, Blockchain-Associated Cognitive Insights Using Blockchains
US20180203744A1 (en) 2017-01-09 2018-07-19 Alexander WIESMAIER Data ingestion and analytics platform with systems, methods and computer program products useful in conjunction therewith
US20190027237A1 (en) 2017-07-21 2019-01-24 Patientory, Inc. Blockchain network for secure exchange of healthcare information
US20190379642A1 (en) * 2018-06-08 2019-12-12 Gcp Ip Holdings I, Llc Blockchain Overwatch

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
GEORGE H.JOHN: "Robust Decision Trees: Removing Outliers from Databases", KDD-95 PROCEEDINGS, pages 174 - 179
HANS-PETER KRIEGELPEER KROGERARTHUR ZIMEK: "Outlier Detection Techniques", SIAM INTERNATIONAL CONFERENCE ON DATA MINING, 2010
K. SIRMANIMANJULA SANJAY KOTI: "Outlier Mining in Medical Databases", STATISTICAL METHODS, vol. 4, no. 01, January 2012 (2012-01-01), pages 239 - 246, ISSN: 0975-5462
MARCO A.F. PIMENTELDAVID A. CLIFTONLEI CLIFTONLIONEL TARASSENKO, A REVIEW OF NOVELTY DETECTION, Retrieved from the Internet <URL:http://dx.doi.Org/10.1016/j.sigpro.2013.12.026>
MARKOS MARKOUSAMEER SINGH: "Novelty Detection: A Review- Part 1: Statistical Approaches", SIGNAL PROCESSING, vol. 83, 2003, pages 2481 - 2497, XP004467982, DOI: 10.1016/j.sigpro.2003.07.018

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20230130086A1 (en) * 2021-10-27 2023-04-27 Arm Limited System, devices and/or processes for modifying electronic contracts

Similar Documents

Publication Publication Date Title
US20230099208A1 (en) Records access and management
EP3583526B1 (fr) Accès à des dossiers et gestion de dossiers
US11580090B2 (en) Community data aggregation with automated followup
EP2365458B1 (fr) Procédé implémenté informatique pour générer un pseudonyme, support de stockage lisible sur ordinateur et système informatique
US20200388359A1 (en) Methods and system of blockchain-based medical service management
US20160359819A1 (en) Encryption and Distribution of Health-related Data
US20180181719A1 (en) Virtual healthcare personal assistant
KR102475987B1 (ko) 인공지능 및 개인건강기록을 이용한 보험정보 제공 시스템 및 그 방법
Oladele et al. BEHeDaS: A Blockchain Electronic Health Data System for Secure Medical Records Exchange
Kaddoura et al. Blockchain for healthcare and medical systems
Poongodi et al. The role of blockchains for medical electronics security
AU2020101898A4 (en) MHOC- Blockchain Technology: Medicine and Healthcare Observation Care using Blockchain Technology
EP3879482A1 (fr) Système et procédés de paiement de soins de santé basé sur la réussite
Al-Ghuraybi et al. Ensuring authentication in Medical Cyber-Physical Systems: a comprehensive literature review of blockchain technology integration with machine learning
Sapra et al. A blockchain solution for the privacy of patients’ medical data
CN117037986A (zh) 医疗或健康记录管理系统及方法
Catanzaro et al. Patients as peers: blockchain based EHR and medical information commons models for HITECH act compliance
Beyari et al. The effect of security information sharing and disruptive technology on patient dissatisfaction in Saudi Health Care Services During Covid-19 pandemic
Zhang Architectures and Patterns for Moving Towards the Use of High-Frequency, Low-Fidelity Data in Healthcare
US20240012934A1 (en) Healthcare-activity triggered data management
Charumilind et al. Ten insights on the US opioid crisis from claims data analysis
Smagulov et al. Digital Transformation of Healthcare
Mao Using Smart and Secret Sharing for Enhanced Authorized Access to Medical Data in Blockchain
Okemiri et al. Patient Data Integration: a panacea for effective healthcare
Savoska et al. Integration of Heterogeneous Data into Electronic Patient Records

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION HAS BEEN PUBLISHED

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE

17P Request for examination filed

Effective date: 20220310

RBV Designated contracting states (corrected)

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: EXAMINATION IS IN PROGRESS

17Q First examination report despatched

Effective date: 20230207